DOCSLIB.ORG

A Gillespie Algorithm for Non-Markovian Stochastic Processes

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
A Gillespie Algorithm for Non-Markovian Stochastic Processes A Gillespie algorithm for non-Markovian stochastic processes 1, 2,3 Naoki Masuda ∗ and Luis E. C. Rocha 1 Department of Engineering Mathematics, University of Bristol, Bristol, UK. 2 Department of Mathematics and naXys, Universit´ede Namur, Namur, Belgium. 3 Department of Public Health Sciences, Karolinska Institutet, Stockholm, Sweden. ∗ Corresponding author ([email protected]) February 20, 2020 Abstract The Gillespie algorithm provides statistically exact methods for simulating stochastic dynamics modelled as interacting sequences of discrete events including systems of biochemical reactions or earthquake occurrences, networks of queuing processes or spiking neurons, and epidemic and opinion formation processes on social networks. Empirically, the inter-event times of vari- ous phenomena obey long-tailed distributions. The Gillespie algorithm and its variants either assume Poisson processes (i.e., exponentially distributed inter-event times), use particular func- tions for time courses of the event rate, or work for non-Poissonian renewal processes, including arXiv:1601.01490v3 [physics.soc-ph] 19 Feb 2020 the case of long-tailed distributions of inter-event times, but at a high computational cost. In the present study, we propose an innovative Gillespie algorithm for renewal processes on the basis of the Laplace transform. The algorithm makes use of the fact that a class of point pro- cesses is represented as a mixture of Poisson processes with different event rates. The method 1 is applicable to multivariate renewal processes whose survival function of inter-event times is completely monotone. It is an exact algorithm and works faster than a recently proposed Gillespie algorithm for general renewal processes, which is exact only in the limit of infinitely many processes. We also propose a method to generate sequences of event times with a tunable amount of positive correlation between inter-event times. We demonstrate our algorithm with exact simulations of epidemic processes on networks, finding that a realistic amount of positive correlation in inter-event times only slightly affects the epidemic dynamics. 1 Introduction Various complex systems are driven by interactions between subsystems via time-stamped dis- crete events. For example, in chemical systems, a chemical reaction event changes the number of reagents of particular types in a discrete manner. Stochastic point processes, in particular Poisson processes assuming that events occur independently and at a constant rate over time, are a central tool for emulating the dynamics of chemical systems [1]. They are also useful in simulating epidemic processes in a population [2] and many other systems. Consider an event-driven system in which events are generated by Poisson processes running in parallel. In chemical reaction systems, each Poisson process (possibly with different rates for each process) is attached to one reaction. In epidemic processes taking place on human or animal contact networks (i.e., graphs), each Poisson process is assigned to an individual or a link, which may potentially transmit the infection. The event rate of some of the Poisson processes may change with the occurrence of a reaction or infection within the entire system. The simplest simulation method is to discretise time and then judge whether or not an event occurs in each time window, for individual processes. This widely used method is suboptimal because the size of the time window must be sufficiently small to obtain high accuracy, which is computationally expensive [3]. The Gillespie algorithm is an efficient and statistically exact 2 algorithm for multivariate Poisson processes [4–6]. The Gillespie algorithm, or, in particular, the direct method of Gillespie [5,6], exploits the fact that a superposition of independent Poisson processes is a single Poisson process whose event rate is the sum of those of the constituent Poisson processes. Using this mathematical property, only a single Poisson process needs to be emulated in the Gillespie algorithm. However, for various real-world systems in which multivariate point processes have been applied (both with and without network structure), event sequences are clearly non-Poissonian. In particular, inter-event times typically obey long-tailed distributions, which are inconsistent with the exponential (i.e., short-tailed) distribution that Poisson processes produce. Examples of long-tailed distributions of inter-event times include earthquake occurrences [7, 8], firing of neurons [9, 10], social networks and the Internet [11–14], financial transactions [15, 16], and crimes [17,18]. Therefore, multivariate point processes that are not necessarily Poissonian have been used in these applications, e.g., triggered seismicity [19, 20], networks of spiking neurons [21, 22], epidemic processes [14, 23], opinion formation models [24, 25], finance [26, 27], and criminology [17,28]. These applications call for numerical methods to efficiently and accurately simulate interacting and non-Markovian point processes. A reasonable description of event sequences in these phenomena requires, at the very least, renewal processes, in which inter-event times are independently generated from a given distri- bution [29, 30]. Along these lines, one numerical approach is to use the modified next reaction method [31], which improves on both the so-called Gillespie’s first reaction method [5] and the next reaction method [32]. The basic idea behind these methods is to draw the next event time for all processes from predetermined distributions, select the process that has generated the minimum waiting time to the next event, execute the event, and repeat. However, for non-Poissonian renewal processes, it is generally difficult to numerically solve the equations that determine the next event time although these methods call for the generation of only half 3 as many random numbers in comparison to the Gillespie algorithm [3]. In addition, we can easily halve the number of random variates required in the Gillespie algorithm, such that the next reaction method and the Gillespie algorithm demand the same number of random variates (Appendix A). In what follows, we restrict ourselves to the Gillespie algorithm and its variants. Motivated by applications to chemical reactions, many extensions of the Gillespie algorithm in the case of non-Poissonian processes assume that the dynamical change in the event rate is exogenously driven in particular functional forms [33, 34]. These extensions are not applicable for general renewal processes, because the event rate of a constituent process is a function of the time since the last event of the process, which depends on that process. In other words, we cannot assume a common exogenous driver. Bogu˜n´aand colleagues extended the Gillespie algorithm to be applicable for general renewal processes [35] (also see [3] for further developments). However, the algorithm has practical limitations. First, it is not accurate when the number of ongoing renewal processes is small [35]. This can result in a considerable amount of approximation error in the beginning or final stages of the dynamics of epidemic or opinion formation models, for example, in which only a small number of processes is active, even in a large population [3]. Second, it is necessary to recalculate the instantaneous event rate of each process following the occurrence of every event in the entire population, a procedure that can be computationally expensive. In the present study, we propose an innovative Gillespie algorithm, the Laplace Gillespie algorithm, which is applicable to non-Poissonian renewal processes. It exploits the mathemati- cal properties of the Laplace transform, is accurate for an arbitrary number of ongoing renewal processes, and runs faster than the previous algorithm [35]. This article is organised as follows. In section 2, we review the original Gillespie algorithm for Poisson processes. In section 3, we review the previous extension of the Gillespie algorithm to general renewal processes [35]. In section 4, we introduce the Laplace Gillespie algorithm, together with theoretical underpin- 4 nings and examples. In section 5, we numerically compare the previous algorithm [35] and the Laplace Gillespie algorithm. In section 6, we introduce a method to generate event sequences with positive correlation in inter-event times, as is typically observed in human behaviour and natural phenomena [13]. In section 7, we demonstrate our methods by performing exact sim- ulations of an epidemic process in which inter-event times follow a power-law distribution. In section 8, we discuss the results, including limitations of the proposed algorithm. The codes used in our numerical simulations are available in the Supplementary Materials. 2 Gillespie algorithm The original Gillespie algorithm [4–6] assumes N independent Poisson processes with a rate of λi (1 i N) running in parallel. Because of the independence of the different Poisson processes, ≤ ≤ N the superposition of the N processes is a Poisson process with rate i=1 λi. Therefore, we first draw ∆t, an increment in time to the next event of the superposedP Poisson process, from the exponential distribution given by N N −(Pi=1 λi)∆t φ(∆t)= λi e . (1) i=1 ! X Because the survival function (i.e., the probability that a random variable is larger than a certain ∞ − N λ ∆t N value) of φ(∆t) is given by φ(t′)dt′ = e (Pi=1 i) , we obtain ∆t = log u λ , ∆t − i=1 i where u is a random variateR drawn from the uniform density on the interval [0, 1]. P Then, we determine the process i that has produced the event with probability λi Πi = N . (2) i=1 λi Finally, we advance the time by ∆t and repeatP the procedure. Following the occurrence of an event, any λi is permitted to change. 5 3 Non-Markovian Gillespie algorithm Now consider N renewal processes running in parallel, and denote by ψi(τ) the probability den- sity function of inter-event times for the ith process (1 i N). If the process is Poissonian, ≤ ≤ −λiτ we obtain ψi(τ)= λie . For such a population of general renewal processes, Bogu˜n´a and col- leagues proposed an extension of the Gillespie algorithm, which they called the non-Markovian Gillespie algorithm (nMGA) [35].
Load more

Recommended publications
  • On Sampling from the Multivariate T Distribution by Marius Hofert
    CONTRIBUTED RESEARCH ARTICLES 129 On Sampling from the Multivariate t Distribution by Marius Hofert Abstract The multivariate normal and the multivariate t distributions belong to the most widely used multivariate distributions in statistics, quantitative risk management, and insurance. In contrast to the multivariate normal distribution, the parameterization of the multivariate t distribution does not correspond to its moments. This, paired with a non-standard implementation in the R package mvtnorm, provides traps for working with the multivariate t distribution. In this paper, common traps are clarified and corresponding recent changes to mvtnorm are presented. Introduction A supposedly simple task in statistics courses and related applications is to generate random variates from a multivariate t distribution in R. When teaching such courses, we found several fallacies one might encounter when sampling multivariate t distributions with the well-known R package mvtnorm; see Genz et al.(2013). These fallacies have recently led to improvements of the package ( ≥ 0.9-9996) which we present in this paper1. To put them in the correct context, we first address the multivariate normal distribution. The multivariate normal distribution The multivariate normal distribution can be defined in various ways, one is with its stochastic represen- tation X = m + AZ, (1) where Z = (Z1, ... , Zk) is a k-dimensional random vector with Zi, i 2 f1, ... , kg, being independent standard normal random variables, A 2 Rd×k is an (d, k)-matrix, and m 2 Rd is the mean vector. The covariance matrix of X is S = AA> and the distribution of X (that is, the d-dimensional multivariate normal distribution) is determined solely by the mean vector m and the covariance matrix S; we can thus write X ∼ Nd(m, S).
    [Show full text]
  • EFFICIENT ESTIMATION and SIMULATION of the TRUNCATED MULTIVARIATE STUDENT-T DISTRIBUTION
    Efficient estimation and simulation of the truncated multivariate student-t distribution Zdravko I. Botev, Pierre l’Ecuyer To cite this version: Zdravko I. Botev, Pierre l’Ecuyer. Efficient estimation and simulation of the truncated multivariate student-t distribution. 2015 Winter Simulation Conference, Dec 2015, Huntington Beach, United States. hal-01240154 HAL Id: hal-01240154 https://hal.inria.fr/hal-01240154 Submitted on 8 Dec 2015 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Proceedings of the 2015 Winter Simulation Conference L. Yilmaz, W. K V. Chan, I. Moon, T. M. K. Roeder, C. Macal, and M. Rosetti, eds. EFFICIENT ESTIMATION AND SIMULATION OF THE TRUNCATED MULTIVARIATE STUDENT-t DISTRIBUTION Zdravko I. Botev Pierre L’Ecuyer School of Mathematics and Statistics DIRO, Universite´ de Montreal The University of New South Wales C.P. 6128, Succ. Centre-Ville Sydney, NSW 2052, AUSTRALIA Montreal´ (Quebec),´ H3C 3J7, CANADA ABSTRACT We propose an exponential tilting method for exact simulation from the truncated multivariate student-t distribution in high dimensions as an alternative to approximate Markov Chain Monte Carlo sampling. The method also allows us to accurately estimate the probability that a random vector with multivariate student-t distribution falls in a convex polytope.
    [Show full text]
  • Stochastic Simulation APPM 7400
    Stochastic Simulation APPM 7400 Lesson 1: Random Number Generators August 27, 2018 Lesson 1: Random Number Generators Stochastic Simulation August27,2018 1/29 Random Numbers What is a random number? Lesson 1: Random Number Generators Stochastic Simulation August27,2018 2/29 Random Numbers What is a random number? “You mean like... 5?” Lesson 1: Random Number Generators Stochastic Simulation August27,2018 2/29 Random Numbers What is a random number? “You mean like... 5?” Formal definition from probability: A random number is a number uniformly distributed between 0 and 1. Lesson 1: Random Number Generators Stochastic Simulation August27,2018 2/29 Random Numbers What is a random number? “You mean like... 5?” Formal definition from probability: A random number is a number uniformly distributed between 0 and 1. (It is a realization of a uniform(0,1) random variable.) Lesson 1: Random Number Generators Stochastic Simulation August27,2018 2/29 Random Numbers Here are some realizations: 0.8763857 0.2607807 0.7060687 0.0826960 Density 0.7569021 . 0 2 4 6 8 10 0.0 0.2 0.4 0.6 0.8 1.0 sample Lesson 1: Random Number Generators Stochastic Simulation August27,2018 3/29 Random Numbers Here are some realizations: 0.8763857 0.2607807 0.7060687 0.0826960 Density 0.7569021 . 0 2 4 6 8 10 0.0 0.2 0.4 0.6 0.8 1.0 sample Lesson 1: Random Number Generators Stochastic Simulation August27,2018 3/29 Random Numbers Here are some realizations: 0.8763857 0.2607807 0.7060687 0.0826960 Density 0.7569021 .
    [Show full text]
  • Multivariate Stochastic Simulation with Subjective Multivariate Normal
    MULTIVARIATE STOCHASTIC SIMULATION WITH SUBJECTIVE MULTIVARIATE NORMAL DISTRIBUTIONS1 Peter J. Ince2 and Joseph Buongiorno3 Abstract.-In many applications of Monte Carlo simulation in forestry or forest products, it may be known that some variables are correlated. However, for simplicity, in most simulations it has been assumed that random variables are independently distributed. This report describes an alternative Monte Carlo simulation technique for subjectivelyassessed multivariate normal distributions. The method requires subjective estimates of the 99-percent confidence interval for the expected value of each random variable and of the partial correlations among the variables. The technique can be used to generate pseudorandom data corresponding to the specified distribution. If the subjective parameters do not yield a positive definite covariance matrix, the technique determines minimal adjustments in variance assumptions needed to restore positive definiteness. The method is validated and then applied to a capital investment simulation for a new papermaking technology.In that example, with ten correlated random variables, no significant difference was detected between multivariate stochastic simulation results and results that ignored the correlation. In general, however, data correlation could affect results of stochastic simulation, as shown by the validation results. INTRODUCTION MONTE CARLO TECHNIQUE Generally, a mathematical model is used in stochastic The Monte Carlo simulation technique utilizes three simulation studies. In addition to randomness, correlation may essential elements: (1) a mathematical model to calculate a exist among the variables or parameters of such models.In the discrete numerical result or outcome as a function of one or case of forest ecosystems, for example, growth can be more discrete variables, (2) a sequence of random (or influenced by correlated variables, such as temperatures and pseudorandom) numbers to represent random probabilities, and precipitations.
    [Show full text]
  • Stochastic Simulation
    Agribusiness Analysis and Forecasting Stochastic Simulation Henry Bryant Texas A&M University Henry Bryant (Texas A&M University) Agribusiness Analysis and Forecasting 1 / 19 Stochastic Simulation In economics we use simulation because we can not experiment on live subjects, a business, or the economy without injury. In other fields they can create an experiment Health sciences they feed (or treat) lots of lab rats on different chemicals to see the results. Animal science researchers feed multiple pens of steers, chickens, cows, etc. on different rations. Engineers run a motor under different controlled situations (temp, RPMs, lubricants, fuel mixes). Vets treat different pens of animals with different meds. Agronomists set up randomized block treatments for a particular seed variety with different fertilizer levels. Henry Bryant (Texas A&M University) Agribusiness Analysis and Forecasting 2 / 19 Probability Distributions Parametric and Non-Parametric Distributions Parametric Dist. have known and well defined parameters that force their shapes to known patterns. Normal Distribution - Mean and Standard Deviation. Uniform - Minimum and Maximum Bernoulli - Probability of true Beta - Alpha, Beta, Minimum, Maximum Non-Parametric Distributions do not have pre-set shapes based on known parameters. The parameters are estimated each time to make the shape of the distribution fit the data. Empirical { Actual Observations and their Probabilities. Henry Bryant (Texas A&M University) Agribusiness Analysis and Forecasting 3 / 19 Typical Problem for Risk Analysis We have a stochastic variable that needs to be included in a business model. For example: Price forecast has residuals we could not explain and they are the stochastic component we need to simulate.
    [Show full text]
  • Stochastic Vs. Deterministic Modeling of Intracellular Viral Kinetics R
    J. theor. Biol. (2002) 218, 309–321 doi:10.1006/yjtbi.3078, available online at http://www.idealibrary.com on Stochastic vs. Deterministic Modeling of Intracellular Viral Kinetics R. Srivastavawz,L.Youw,J.Summersy and J.Yinnw wDepartment of Chemical Engineering, University of Wisconsin, 3633 Engineering Hall, 1415 Engineering Drive, Madison, WI 53706, U.S.A., zMcArdle Laboratory for Cancer Research, University of Wisconsin Medical School, Madison, WI 53706, U.S.A. and yDepartment of Molecular Genetics and Microbiology, University of New Mexico School of Medicine, Albuquerque, NM 87131, U.S.A. (Received on 5 February 2002, Accepted in revised form on 3 May 2002) Within its host cell, a complex coupling of transcription, translation, genome replication, assembly, and virus release processes determines the growth rate of a virus. Mathematical models that account for these processes can provide insights into the understanding as to how the overall growth cycle depends on its constituent reactions. Deterministic models based on ordinary differential equations can capture essential relationships among virus constituents. However, an infection may be initiated by a single virus particle that delivers its genome, a single molecule of DNA or RNA, to its host cell. Under such conditions, a stochastic model that allows for inherent fluctuations in the levels of viral constituents may yield qualitatively different behavior. To compare modeling approaches, we developed a simple model of the intracellular kinetics of a generic virus, which could be implemented deterministically or stochastically. The model accounted for reactions that synthesized and depleted viral nucleic acids and structural proteins. Linear stability analysis of the deterministic model showed the existence of two nodes, one stable and one unstable.
    [Show full text]
  • Random Numbers and Stochastic Simulation
    Stochastic Simulation and Randomness Random Number Generators Quasi-Random Sequences Scientific Computing: An Introductory Survey Chapter 13 – Random Numbers and Stochastic Simulation Prof. Michael T. Heath Department of Computer Science University of Illinois at Urbana-Champaign Copyright c 2002. Reproduction permitted for noncommercial, educational use only. Michael T. Heath Scientific Computing 1 / 17 Stochastic Simulation and Randomness Random Number Generators Quasi-Random Sequences Stochastic Simulation Stochastic simulation mimics or replicates behavior of system by exploiting randomness to obtain statistical sample of possible outcomes Because of randomness involved, simulation methods are also known as Monte Carlo methods Such methods are useful for studying Nondeterministic (stochastic) processes Deterministic systems that are too complicated to model analytically Deterministic problems whose high dimensionality makes standard discretizations infeasible (e.g., Monte Carlo integration) < interactive example > < interactive example > Michael T. Heath Scientific Computing 2 / 17 Stochastic Simulation and Randomness Random Number Generators Quasi-Random Sequences Stochastic Simulation, continued Two main requirements for using stochastic simulation methods are Knowledge of relevant probability distributions Supply of random numbers for making random choices Knowledge of relevant probability distributions depends on theoretical or empirical information about physical system being simulated By simulating large number of trials, probability
    [Show full text]
  • Arxiv:2008.11456V2 [Physics.Optics]
    Efficient stochastic simulation of rate equations and photon statistics of nanolasers Emil C. Andr´e,1 Jesper Mørk,1, 2 and Martijn Wubs1,2, ∗ 1Department of Photonics Engineering, Technical University of Denmark, Ørsteds Plads 345A, DK-2800 Kgs. Lyngby, Denmark 2NanoPhoton - Center for Nanophotonics, Technical University of Denmark, Ørsteds Plads 345A, DK-2800 Kgs. Lyngby, Denmark Based on a rate equation model for single-mode two-level lasers, two algorithms for stochastically simulating the dynamics and steady-state behaviour of micro- and nanolasers are described in detail. Both methods lead to steady-state photon numbers and statistics characteristic of lasers, but one of the algorithms is shown to be significantly more efficient. This algorithm, known as Gillespie’s First Reaction Method (FRM), gives up to a thousandfold reduction in computation time compared to earlier algorithms, while also circumventing numerical issues regarding time-increment size and ordering of events. The FRM is used to examine intra-cavity photon distributions, and it is found that the numerical results follow the analytics exactly. Finally, the FRM is applied to a set of slightly altered rate equations, and it is shown that both the analytical and numerical results exhibit features that are typically associated with the presence of strong inter-emitter correlations in nanolasers. I. INTRODUCTION transitions from thermal to coherent emission, and we will investigate the prospects of introducing effectively Optical cavities on the micro- and nanometer scale can altered rates of emission in the rate equations as a way reduce the number of available modes for light emission to qualitatively describe the effects of collective inter- and increase the coupling of spontaneously emitted light emitter correlations.
    [Show full text]
  • Simulation and the Monte Carlo Method
    SIMULATION AND THE MONTE CARLO METHOD Third Edition Reuven Y. Rubinstein Technion Dirk P. Kroese University of Queensland Copyright © 2017 by John Wiley & Sons, Inc. All rights reserved. Published by John Wiley & Sons, Inc., Hoboken, New Jersey. Published simultaneously in Canada. Library of Congress Cataloging-in-Publication Data: Names: Rubinstein, Reuven Y. | Kroese, Dirk P. Title: Simulation and the Monte Carlo method. Description: Third edition / Reuven Rubinstein, Dirk P. Kroese. | Hoboken, New Jersey : John Wiley & Sons, Inc., [2017] | Series: Wiley series in probability and statistics | Includes bibliographical references and index. Identifiers: LCCN 2016020742 (print) | LCCN 2016023293 (ebook) | ISBN 9781118632161 (cloth) | ISBN 9781118632208 (pdf) | ISBN 9781118632383 (epub) Subjects: LCSH: Monte Carlo method. | Digital computer simulation. | Mathematical statistics. | Sampling (Statistics) Classification: LCC QA298 .R8 2017 (print) | LCC QA298 (ebook) | DDC 518/.282--dc23 LC record available at https://lccn.loc.gov/2016020742 Printed in the United States of America CONTENTS Preface xiii Acknowledgments xvii 1 Preliminaries 1 1.1 Introduction 1 1.2 Random Experiments 1 1.3 Conditional Probability and Independence 2 1.4 Random Variables and Probability Distributions 4 1.5 Some Important Distributions 5 1.6 Expectation 6 1.7 Joint Distributions 7 1.8 Functions of Random Variables 11 1.8.1 Linear Transformations 12 1.8.2 General Transformations 13 1.9 Transforms 14 1.10 Jointly Normal Random Variables 15 1.11 Limit Theorems 16
    [Show full text]
  • A Deterministic Monte Carlo Simulation Framework for Dam Safety Flow Control Assessment
    water Article A Deterministic Monte Carlo Simulation Framework for Dam Safety Flow Control Assessment Leanna M. King * and Slobodan P. Simonovic Department of Civil and Environmental Engineering, Western University, London, ON N6A3K7, Canada; [email protected] * Correspondence: [email protected] Received: 17 December 2019; Accepted: 5 February 2020; Published: 12 February 2020 Abstract: Simulation has become more widely applied for analysis of dam safety flow control in recent years. Stochastic simulation has proven to be a useful tool that allows for easy estimation of the overall probability of dam overtopping failure. However, it is difficult to analyze “uncommon combinations of events” with a stochastic approach given current computing abilities, because (a) the likelihood of these combinations of events is small, and (b) there may not be enough simulated instances of these rare scenarios to determine their criticality. In this research, a Deterministic Monte Carlo approach is presented, which uses an exhaustive list of possible combinations of events (scenarios) as a deterministic input. System dynamics simulation is used to model the dam system interactions so that low-level events within the system can be propagated through the model to determine high-level system outcomes. Monte Carlo iterations are performed for each input scenario. A case study is presented with results from a single example scenario to demonstrate how the simulation framework can be used to estimate the criticality parameters for each combination of events simulated. The approach can analyze these rare events in a thorough and systematic way, providing a better coverage of the possibility space as well as valuable insights into system vulnerabilities.
    [Show full text]
  • Stochastic Simulation Script for the Course in Spring 2012
    Stochastic Simulation Script for the course in spring 2012 Hansruedi K¨unsch Seminar f¨urStatistik, ETH Zurich June 2012 Contents 1 Introduction and Examples 1 1.1 What is Stochastic Simulation? . 1 1.2 Distribution of Estimators and Test Statistics . 2 1.2.1 Precision of the trimmed mean . 2 1.2.2 Bootstrap . 4 1.3 Simulation in Bayesian Statistics . 5 1.3.1 Conditional distributions of continuous random vectors . 5 1.3.2 Introduction to Bayesian statistics . 6 1.4 Simulations in Statistical Mechanics . 8 1.5 Simulations in Operations Research . 10 1.6 Simulations in Financial Mathematics . 10 1.7 The accuracy of Monte Carlo methods . 12 1.8 Other applications of stochastic simulation . 14 2 Generating Uniform Random Variables 15 2.1 Linear Congruential Generators . 16 2.2 Other Generators . 18 2.3 Combination of Generators . 19 2.4 Testing of Random Numbers . 20 3 Direct Generation of Random Variables 23 3.1 Quantile Transform . 23 3.2 Rejection Sampling . 24 3.3 Ratio of Uniform Random Variables . 26 3.4 Relations between Distributions . 27 3.4.1 The Normal Distribution . 27 3.4.2 The Poisson Distribution . 29 3.5 Summary: Simulation of Selected Distributions . 29 3.6 Random Sampling and Random Permutations . 31 3.7 Importance Sampling . 32 3.8 Markov Chains and Markov Processes . 32 3.8.1 Simulation of Stochastic Differential Equations . 33 3.9 Variance Reduction . 35 3.9.1 Antithetic Variates . 35 3.9.2 Control Variates . 36 3.9.3 Importance Sampling and Variance Reduction . 37 3.9.4 Quasi-Monte Carlo .
    [Show full text]
  • Statistics and Simulation
    Statistics and Simulation By Smita Skrivanek, Principal Statistician, MoreSteam.com LLC Stochastic Simulation Stochastic simulation has played an increasingly large role in statistics with rapid increases in computing power. The exercises in the MoreSteam courses are examples of simulation. Simulations are used to model complicated processes, estimate distributions of estimators (using methods such as bootstrap), and have dramatically increased the use of an entire field of statistics, Bayesian statistics, since simulation can be used to estimate the distribution of its statistics, which are usually analytically intractable except in the simplest situations. So what is ‘stochastic simulation’? Stochastic simulation, a.k.a Monte Carlo simulation, is the use of random number generators to produce data from a mathematical model. A random number generator is any process that produces data whose observations are independent and identically distributed (i.i.d) from some distribution. It is a method used to evaluate processes or systems using probabilities based on sampling repeatedly from randomly generated numbers. Simulation by random sampling was in use during the genesis of probability theory, long before computers. Toss a coin a hundred times and count the number of times it falls Heads. Dividing the total number of times the coin falls Heads by the total number of tosses gives you an estimate of the probability of Heads for that coin, using simulation. The error associated with that estimate is called “Monte Carlo error”. Stochastic simulation was used as far back as in 1777 by Buffon to estimate the probability of a needle falling across a line in a uniform grid of parallel lines, known as Buffon’s needle problem.
    [Show full text]